Did Gregorian chant descend from Old Roman chant, or did both come from a common ancestor?

Frankish 'Gregorian' and city-of-Rome 'Old Roman' chant share texts and melodic outlines but differ in detail, and Roman melodies were written down only from 1071.

Texts & authorship · Europe · 680–1250

Could use help from musicologists, statisticians and data scientists and philologists and epigraphers.

Neumed introit Omnes gentes, Einsiedeln Codex 121, later tenth century. Unknown author (scan via e-codices), Public domain

Recent work

last moved Jun 2026

2016’162017’172018’182019’192020’202021’212022’222023’232024’242025’252026’26

Jun 2026Tool
Towards a foundational model for recognising diastematic Gregorian notationD. Kurek and J. Hajič jr · arXivUnifies four optical music recognition datasets for pitched chant notation in one encoding and trains a shared model, a route to transcribing pitched sources at scale.
May 2026Paper
What did the dove sing to Pope Gregory? Ancestral melody reconstruction in Gregorian chant using Bayesian phylogeneticsG. A. Ballen, K. H. Mühlová and J. Hajič jr · PLOS ONEReconstructs ancestral melodies on a time-calibrated tree of fourteen sources and proposes encoding Old Roman sources to test their chronology against Gregorian.Advances the problem
Apr 2026Data
Making Chant Computing Easy: CantusCorpus v1.0 and the PyCantus LibraryA. Dvořáková, T. Eipert, D. Lacoste and J. Hajič jr · Transactions of the International Society for Music Information RetrievalReleases a single dataset combining the Cantus Database and allied Cantus Index databases (nearly 900,000 chants in over 2,000 sources) with a Python library.
2026Paper
A variánsalkotás mintázatai órómai és gregorián kéziratokban (Patterns of variants in Old Roman and Gregorian chant manuscripts)L. D. Varga · Magyar Zene 64 (1)Applies digital stemmatics and descriptive statistics to melodic variants in the Old Roman sources, read against their Gregorian analogues, to infer sub-traditions among the Roman books.
Dec 2025Paper
Genome of melody: applying bioinformatics to study the evolution of Gregorian chantJ. Hajič jr, V. Lanz and G. A. Ballen · Philosophical Transactions of the Royal Society BShows that phylogenetic models on aligned melodies recover historically plausible patterns among Gregorian sources; Old Roman sources are not included.
Jul 2025Paper
Gregorian melody, modality, and memory: Segmenting chant with Bayesian nonparametricsV. Lanz and J. Hajič jr · arXivTests the centonisation idea that chant is built from reusable melodic segments, relevant to separating formulaic convergence from shared descent.
Jan 2023Tool
ChantLab and Genome of Melody code repositoriesGenome of Melody project (J. Hajič jr et al.) · GitHubOpen code for aligning volpiano chant melodies and running MrBayes phylogenetic analyses, released alongside the group's chant phylogeny papers.
Apr 2022Paper
Revisiting 'Toledo, Rome, and the Legacy of Gaul': new evidence from the Divine OfficeR. Maloy et al. · Plainsong and Medieval MusicIdentifies nearly two dozen Franco-Roman responsories shared with the Old Hispanic rite, extending Levy's case for non-Roman material in the Gregorian repertory.
Jun 2021Paper
Renewal, Heritage, and Exchange in Eleventh-Century Roman Chant TraditionsL. Nardini · Urban Developments in Late Antique and Medieval Rome (Amsterdam UP)Examines the eleventh-century Roman chant books, the period when Old Roman melodies were first written down, as products of local renewal and exchange.
Oct 2020Tool
Studying Large Plainchant Corpora Using chant21B. Cornelissen, W. Zuidema and J. A. Burgoyne · DLfM 2020Python library for parsing Cantus volpiano and GABC chant encodings into music21, making corpus-scale melodic analysis practicable.
Nov 2018Paper
Writing Sounds in Carolingian Europe: The Invention of Musical NotationS. Rankin · Cambridge University PressReconstructs the ninth-century emergence of neume notation, constraining when Frankish melodies could first have been fixed in writing.
Oct 2017Paper
Comparing Offertory Melodies of Five Medieval Christian Chant TraditionsP. van Kranenburg and G. Maessen · ISMIR 2017 ProceedingsApplies computational melodic comparison to offertories shared across five Latin chant traditions, including the Old Roman and Gregorian versions.
Oct 2017Paper
The Earliest Antiphons of the Roman OfficeE. Nowacki · Chant, Liturgy, and the Inheritance of Rome (Boydell)Uses Old Roman and Gregorian Office antiphons comparatively to identify the oldest layer of the Roman Office, in a volume of essays honouring Joseph Dyer.

Compiled by Claude Opus 5.5, 2026-10-11

Why it matters

Gregorian chant is the foundation of Western music writing, modal theory and the later polyphonic tradition, and from the ninth century it was presented as the authentic chant of Rome. Whether it records the Roman repertory imported under Pippin III and Charlemagne, a Frankish re-composition of it, or a fusion with Gallican chant decides what the Carolingian 'romanization' of liturgy actually involved. Because the Roman melodies were written down roughly three centuries after the transfer, the answer is also a test case for how faithfully unnotated melody can be transmitted, which bears on oral-tradition arguments well beyond chant.

Why it is open

The decisive period is undocumented in pitch. The Frankish chant books of about 800 have texts only; the neumed graduals of about 900 to 970, such as , and , record melodic shape but not exact pitch; and Roman melodies survive only from (1071) and a handful of later books. The same differences can be explained in opposite directions: the Franks may have simplified an ornate Roman style, the Romans may have drifted into formulaic uniformity during three centuries of oral transmission, or the Romans may have reworked imported Gregorian melodies in their own idiom. Comparison has proceeded genre by genre and chant by chant with qualitative judgement, and most Old Roman melodies are still not encoded in machine-readable form.

What would count as a solution

A pre-registered comparison of rival transmission models (Old Roman ancestral to Gregorian, independent descent from a common eighth-century Roman tradition, and Roman reworking of imported Gregorian chant), fitted to aligned melodies of all shared Mass propers and scored by how well each predicts held-out readings in early pitched and neumed sources, with decision margins fixed before the test. A model that wins across genres by the agreed margin settles the question for practical purposes; a result that differs by genre is also an answer, showing that layers of the repertory have different histories.

Potential approach

The comparative literature is the starting point: (1956) framed the 'central problem', (1970) edited , (1974) and (1980) recast both repertories as products of oral transmission, and genre studies by (1998) on offertories, (2002) on eighth-mode tracts and (2010) on offertories supply chant-by-chant judgements about which readings look older.
The rival positions should be stated as testable predictions before any data are fitted: Roman priority in a Frankish adaptation, as in Hucke (1980) and (2000); Old Roman as a later, formulaic development of the common tradition, as in (2002) and Maloy (2010); and (2003), in which the Franks drew on Gallican material and the Romans later absorbed and remodelled Gregorian melodies. Each predicts a different direction of change for ornament, cadence formulas and melodic range.
Teams could encode every Old Roman Mass proper in , Vat. lat. 5319 and (and Office chants in and ) in volpiano, link them to Gregorian melodies already in the through Cantus Index identifiers, and add the neume contours of , and as partial observations of pitch direction.
The phylogenetic methods of Hajič, Lanz and (2025) and the ancestral melody reconstruction of Ballen, Mühlová and Hajič (2026), which so far cover Gregorian sources only, can be extended to the two repertories, with care taken that formulaic convergence (shared stock cadences) is not mistaken for shared descent; (1997) is an earlier computer-assisted study of Old Roman formulas to build on.
Held-out testing gives the decision: hide a set of early pitched witnesses and the neumed contours of the earliest Frankish graduals, and score each model on how well it predicts them, with thresholds, genres and alignment rules published in advance so that the result cannot be tuned after the fact.
Text evidence can serve as an independent check: the psalter forms and biblical variants in shared chant texts, which Pfisterer (2002) used, date the wording of each repertory without depending on the melodies.

Drafted by Claude Opus 5.5, 2026-10-11

Who can help

musicologists: Encode the Old Roman Mass propers of Bodmer 74, Vat. lat. 5319 and San Pietro F 22 in volpiano, aligned chant by chant with Gregorian sources through Cantus Index identifiers.
statisticians and data scientists: Pre-register rival transmission models (Roman-derived, common ancestor, later Roman reworking of Gregorian) and score each by how well it predicts held-out readings in early pitched sources.
philologists and epigraphers: Classify the psalter and biblical text variants in shared chant texts as Roman, Gallican or Vulgate forms, as independent evidence for which repertory kept the older wording.